EECS 461, Winter 2009, Problem Set 21

issued: Wednesday, January 28, 2009 due: Wednesday, February 4, 2009.

1. In all sensor interfacing, it is necessary to minimize the response of the system to noise in the measure-
ments. For example, in quadrature decoding, noise can cause spurious pulses. These can lead to erroneous
updates of the counter, and thus to erroneous position measurements.
There is a certain amount of noise immunity built into the MPC5553 eTPU, and thus into the QD function
we use in lab. This is explained in Section 5.8.6 of the eTPU Reference Manual, available on the class
website. Consider so-called “two sample mode.” There is a digital filter that samples the input two times
with a sample period that is a multiple of the system clock, which in our case is set to 40MHz. If the
multiplier is equal to 2, then the sample frequency is 20MHz. If the samples are identical, then the input
signal state is updated. A pulse that lasts only one sample is treated as noise. Hence we see that

(i). Pulses of 2 CPU clocks or less in duration are rejected.

(a). What is the signal latency contributed by noise suppression to the QD function?
(b). How does the latency change if the filter multiplier is set to 4?
(c). Suppose the filter multiplier is set to 2, and a spurious pulse of finite duration greater than 4 CPU
clocks appears on only one channel. Does this cause a permanent error in position count? Explain.
(d). Suppose there are overlapping spurious pulses on both channels. Does this change your answer from
question (c) above?

2. In this problem, we will use Matlab, Simulink, and Stateflow to construct a simplified model of a quadra-
ture decoding algorithm. The purpose of the problem is both to learn more about quadrature decoding
as well as to learn how to set up and run a Stateflow model. Much embedded control software consists
of event driven state transitions, and is best modelled using finite state machines. In fact, Stateflow was
developed so that engineers could model the interaction between such software and a model of a dynam-
ical system. Stateflow uses Harel statecharts, which are also popular in the object oriented programming
community.
There are several Matlab files available on the class web site. The first of these, “run quad decode.m”, is
used to generate the two quadrature signals shown in Figure 1. These signals are then used as inputs to
the Simulink model “quad decode fast.mdl”, shown in Figures 2-3, written to model the “fast” mode of
quadrature decoding as implemented on the MPC5553. Note several features of this model:
(i). the secondary channel is ignored.
(ii). by clicking on the Tools→Explore menu of the state chart, one can see that a transition is triggered
only on the rising edges of the primary channel
(iii). on the same menu, one sees that the counter is initialized to zero
(iv). the counter is incremented in multiples of 4 (why?)
Run the Matlab file “run quad decode.m”, and inspect the plot of counter value vs time.

(a) What value does the counter have at the end of the simulation?
(b) Are the changes in counter value you see consistent with your understanding of the fast quadrature
decode mode of the MPC5553?
(c) Does the final value of the counter correctly represent the change in wheel position?
1 Revised January 22, 2009

1
 1.5

Channel 1 (primary)
1

0.5

−0.5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

1.5
Channel 2 (secondary)

0.5

−0.5
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
time, T units

Figure 1: Quadrature Signals from Encoder

(d) Now inspect Figures 4-5, which are a partial implementation of the slow decode mode of the MPC5553.
Change the Simulink model to have the form shown in Figure 4. To do so, open the “Mux” block and
change the number of inputs to 2, remove the terminator from the secondary input, and enter it into
the “Mux” block. With both primary and secondary inputs available, one can perform slow mode,
or 4X, quadrature decoding by using a state machine with 4 states, as shown in Figure 5. Complete
Figure 5 by adding transitions between the four states. You will need to open the Tools→Explore
menu of your Stateflow block, add the secondary input as an input event, and set the two events to
trigger appropriately.
Your completed model of normal mode quadrature decoding should yield the plot of counter value
shown in Figure 6. Hand in the corresponding plot from your program to show that it achieves the
same result, as well as your Stateflow chart. (We reserve the right to test your program on a different
set of quadrature signals, so you should perform such tests yourself to make sure that your code works
in general, and not merely for the given examples.)

2
 2
input1

Channel1

fqd1
From 0
Workspace

Counter

fqd2
From Terminator
Workspace1
counter 1
Channel2 Counter1

3 Chart
input2

output

Figure 2: Simulink Model for Fast Mode Quadrature Decoding

quad_decode_fast/Chart

primary{counter=counter+4;}
state0

Printed 18−Sep−2002 18:27:52

Figure 3: Stateflow Chart for Fast Mode Quadrature Decoding

3
 2
input1
Channel1

fqd1
0
From
Workspace
Counter

fqd2
From
Workspace1
3 counter 1
input2 Counter1

Chart
Channel2

output

Figure 4: Simulink Model for Normal (4X) Mode Quadrature Decoding

quad_decode_4X_skeleton/Chart

state00 state10

state01 state11

Printed 18−Sep−2002 21:34:05

Figure 5: Incomplete Stateflow Chart for Normal (4X) Mode Quadrature Decoding

4
 15

counter value
10

0
Channel 1 (primary) 0 5 10 15 20 25

0.5

0 5 10 15 20 25
Channel 2 (secondary)

0.5

0 5 10 15 20 25
time, T units

Figure 6: Counter Value vs Time for Normal Mode Decoding